Event Sequence Pattern Mining for Low-Noise Trie Visualization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contact center event sequences exhibit significant noise and complexity, leading to interpretability and explainability issues, particularly in visualization, and the underlying data structures like Trie data structures become excessively large and complex, resulting in poor performance and higher operational costs.
Innovation Solution
Implement pattern mining to identify common patterns in event sequences, encode them into vector embeddings, apply clustering algorithms to reduce noise, and build a Trie data structure for event sequence visualization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If event sequences are visualized using traditional Trie data structures, then complete event sequence information is preserved, but the data structure becomes excessively large and complex leading to poor performance
Solution Approach 1:
The patent extracts and removes noise events from the event sequences before building the Trie data structure. By identifying and eliminating irrelevant or redundant events, the data structure size and complexity are reduced while preserving the essential event sequence information needed for analysis and visualization.
Solution Approach 2:
The patent applies parameter changes by modifying the event sequence data through noise reduction techniques before Trie construction. This preprocessing step changes the parameters of the event sequences (removing noise events), which directly reduces the resulting Trie data structure's size and complexity while maintaining its informational value.
2Loss of information
If all event sequences are processed and visualized, then complete analysis coverage is achieved, but noise and complexity increase leading to interpretability issues
Solution Approach 1:
The patent extracts meaningful event patterns from noisy event sequences by removing noise events. This extraction process maintains comprehensive analysis coverage by preserving all significant events while eliminating noise, thereby improving the interpretability of the visualized event sequences without losing important information.
Solution Approach 2:
The patent segments the event sequences by identifying and separating noise events from meaningful events. This segmentation allows the system to process and visualize only the relevant event patterns, maintaining complete analysis coverage of meaningful events while improving interpretability by excluding noisy data.
3Loss of information
If traditional Trie data structures are used for event sequence visualization, then all event details are retained, but computational costs and operational expenses increase
Solution Approach 1:
The patent extracts and removes noise events from the event sequences before Trie construction, which reduces the amount of data that needs to be processed, stored, and visualized. This extraction maintains retention of all meaningful event details while significantly reducing computational costs and operational expenses associated with handling the full event dataset.
Solution Approach 2:
The patent applies parameter changes by preprocessing event sequences to remove noise events, which reduces the data volume parameter. This parameter change leads to lower computational costs for Trie construction, maintenance, and query operations, while preserving all essential event detail information needed for accurate analysis.
Data Source
AI summary
A method for reducing noise for event sequence visualization according to an embodiment includes identifying patterns of events from an event sequence dataset, wherein the event sequence dataset includes data for a plurality of event sequences, and wherein each event sequence of the plurality of events sequences includes at least one event, encoding each event sequence of the plurality of event sequences into a respective vector embedding based on the identified patterns of events to generate a plurality of vectors, executing a clustering algorithm on the plurality of vectors to generate a plurality of clusters, assigning each event sequence of the plurality of event sequences to a respective cluster of the plurality of clusters, generating a reduced dataset based on the assignment of the plurality of event sequences to the plurality of clusters, and building a data structure for event sequence visualization based on the reduced dataset.